Redis 的坑

redisTemplate封装指令和redis connect原生指令混用导致json解析失败

redis connect原生指令在存储值时要注意转json格式，单行字符串不转json要记得两边加双引号，否则redisTemplate取数据json转对象时会解析失败。

存Long取出时是Integer及类似问题

项目中redis序列化配置组件统一使用

Jackson2JsonRedisSerializer jackson2JsonRedisSerializer =newJackson2JsonRedisSerializer(Object.class);

Map<String, Long> objectMaps = redisTemplate.<String, Long>opsForHash().entries("hashkey");
spaceId = objectMaps.get(key)

会出BUG

 java.lang.ClassCastException: java.lang.Integer cannot be cast to java.lang.String

一步一步看源码(DEBUG过程中不涉及到的代码会删除)
首先将redis中数据转成转成Map<byte[], byte[]>，然后反序列化。

class DefaultHashOperations<K, HK, HV> extends AbstractOperations<K, Object> implements HashOperations<K, HK, HV> {

DefaultHashOperations(RedisTemplate<K, ?> template) {
   super((RedisTemplate<K, Object>) template);
}

AbstractOperations(RedisTemplate<K, V> template) {
   this.template = template;
}

public Map<HK, HV> entries(K key) {

   byte[] rawKey = rawKey(key);
   Map<byte[], byte[]> entries = execute(connection -> connection.hGetAll(rawKey));

   return entries != null ? deserializeHashMap(entries) : Collections.emptyMap();
}
}

反序列化代码
调用反序列代码时没有指定<HK, HV>，等效于返回了一个Map<>,JVM在编译时擦除泛型，HK, HV会替换成Object
问题出在下面的代码。
举个例子
下面一段代码（称为代码one）会绕过类型检查，不会抛出异常和警告，其中类A是上面class DefaultHashOperations的简写，类B是下面反序列化代码的简写，代码one这种编写方式使得A类调用B类时不会发出异常和警告，只会在Map拿出来数据时出现类型转换错误。

class  A {
   B=new B();
    @Test
    public void aa() {
        Map<String,Long> map=i.returnMap();
    }
}
class B{
    public<K,V> Map<K,V> returnMap(){
        return new HashMap<>();
    }
}

下面一段代码只会给出警告，在取值时会抛出异常。
但为什么在使用redis api时没有抛出警告？
首先，类似于代码one的编写方式不会有任何异常和警告，完全越过了类型检查。
其次，反序列化中很多方法上使用了@SuppressWarnings("unchecked")注解，这个注解会抑制警告，不是很理解为什么要抑制警告，很差劲的编写习惯。

HashMap hashMap=new HashMap<>();
hashMap.put(1521,"afdsdfa");
Map<String, Long> objectMaps = hashMap;

abstract class AbstractOperations<K, V> {
    @SuppressWarnings("unchecked")
    <HK, HV> Map<HK, HV> deserializeHashMap(@Nullable Map<byte[], byte[]> entries) {
        // connection in pipeline/multi mode
        if (entries == null) {
            return null;
        }

        Map<HK, HV> map = new LinkedHashMap<>(entries.size());

        for (Map.Entry<byte[], byte[]> entry : entries.entrySet()) {
            map.put((HK) deserializeHashKey(entry.getKey()), (HV) deserializeHashValue(entry.getValue()));
        }

        return map;
    }
 @SuppressWarnings("unchecked")
    <HK> HK deserializeHashKey(byte[] value) {
        return (HK) hashKeySerializer().deserialize(value);
    }
@SuppressWarnings("unchecked")
    <HV> HV deserializeHashValue(byte[] value) {
        return (HV) hashValueSerializer().deserialize(value);
    }

RedisSerializer hashKeySerializer() {
   return template.getHashKeySerializer();
}

RedisSerializer hashValueSerializer() {
   return template.getHashValueSerializer();
}
}

Jackson2JsonRedisSerializer中序列化类型在上文项目配置构造方法中指定为 Object.class ，除此之外提供没有别的途径指定反序列化后得到的类型。

public class Jackson2JsonRedisSerializer<T> implements RedisSerializer<T> {

private final JavaType javaType;
public Jackson2JsonRedisSerializer(Class<T> type) {
   this.javaType = getJavaType(type);
}
@SuppressWarnings("unchecked")
public T deserialize(@Nullable byte[] bytes) throws SerializationException {
      return (T) this.objectMapper.readValue(bytes, 0, bytes.length, javaType);
}
}

public class ObjectMapper extends ObjectCodec implements Versioned,java.io.Serializable 
{

public <T> T readValue(byte[] src, int offset, int len, JavaType valueType)
    throws IOException, StreamReadException, DatabindException
{
    return (T) _readMapAndClose(_jsonFactory.createParser(src, offset, len), valueType);
} 

protected Object _readMapAndClose(JsonParser p0, JavaType valueType)
    throws IOException
{
    try (JsonParser p = p0) {
        final Object result;
        final DeserializationConfig cfg = getDeserializationConfig();
        final DefaultDeserializationContext ctxt = createDeserializationContext(p, cfg);
        JsonToken t = _initForReading(p, valueType);
        if (t == JsonToken.VALUE_NULL) {
        
        } else if (t == JsonToken.END_ARRAY || t == JsonToken.END_OBJECT) {
        
        } else {
            result = ctxt.readRootValue(p, valueType,_findRootDeserializer(ctxt, valueType), null);
        }
        return result;
    }
  

}

public abstract class DefaultDeserializationContext extends DeserializationContext implements java.io.Serializable{
public Object readRootValue(JsonParser p, JavaType valueType,
        JsonDeserializer<Object> deser, Object valueToUpdate)
    throws IOException
{
    if (valueToUpdate == null) {
        return deser.deserialize(p, this);
    }
    return deser.deserialize(p, this, valueToUpdate);
}
}

到这一步，涉及变量类型的只有最初传进来的Obejct.class和下面构造方法定义的四种基本类型，再往下的代码看不懂，但基本可以断定如果没有指定特定类型就会转化为Boolean，Integer，Double,String。

public class AbstractDeserializer extends JsonDeserializer<Object> implements ContextualDeserializer,  java.io.Serializable
{

private static final long serialVersionUID = 1L;

protected final JavaType _baseType;

protected final ObjectIdReader _objectIdReader;

protected final Map<String, SettableBeanProperty> _backRefProperties;

protected transient Map<String,SettableBeanProperty> _properties;

// support for "native" types, which require special care:

protected final boolean _acceptString;
protected final boolean _acceptBoolean;
protected final boolean _acceptInt;
protected final boolean _acceptDouble;

public AbstractDeserializer(BeanDeserializerBuilder builder,
        BeanDescription beanDesc, Map<String, SettableBeanProperty> backRefProps,
        Map<String, SettableBeanProperty> props)
{
    _baseType = beanDesc.getType();
    _objectIdReader = builder.getObjectIdReader();
    _backRefProperties = backRefProps;
    _properties = props;
    Class<?> cls = _baseType.getRawClass();
    _acceptString = cls.isAssignableFrom(String.class);
    _acceptBoolean = (cls == Boolean.TYPE) || cls.isAssignableFrom(Boolean.class);
    _acceptInt = (cls == Integer.TYPE) || cls.isAssignableFrom(Integer.class);
    _acceptDouble = (cls == Double.TYPE) || cls.isAssignableFrom(Double.class);
}

@Override
public Object deserialize(JsonParser p, DeserializationContext ctxt)
    throws IOException
{
    // 16-Oct-2016, tatu: Let's pass non-null value instantiator so that we will
    //    get proper exception type; needed to establish there are no creators
    //    (since without ValueInstantiator this would not be known for certain)
    ValueInstantiator bogus = new ValueInstantiator.Base(_baseType);
    return ctxt.handleMissingInstantiator(_baseType.getRawClass(), bogus, p,
            "abstract types either need to be mapped to concrete types, have custom deserializer, or contain additional type information");
}

这是最底层，但我现在看不懂…

public abstract class DeserializationContext
    extends DatabindContext
    implements java.io.Serializable
{

public Object handleMissingInstantiator(Class<?> instClass, ValueInstantiator valueInst,
            JsonParser p, String msg, Object... msgArgs)
        throws IOException
    {
        if (p == null) {
            p = getParser();
        }
        msg = _format(msg, msgArgs);
        LinkedNode<DeserializationProblemHandler> h = _config.getProblemHandlers();
        while (h != null) {
            // Can bail out if it's handled
            Object instance = h.value().handleMissingInstantiator(this,
                    instClass, valueInst, p, msg);
            if (instance != DeserializationProblemHandler.NOT_HANDLED) {
                // Sanity check for broken handlers, otherwise nasty to debug:
                if (_isCompatible(instClass, instance)) {
                    return instance;
                }
                reportBadDefinition(constructType(instClass), String.format(
"DeserializationProblemHandler.handleMissingInstantiator() for type %s returned value of type %s",
                    ClassUtil.getClassDescription(instClass),
                    ClassUtil.getClassDescription((instance)
                )));
            }
            h = h.next();
        }

        // 16-Oct-2016, tatu: This is either a definition problem (if no applicable creator
        //   exists), or input mismatch problem (otherwise) since none of existing creators
        //   match with token.
        // 24-Oct-2019, tatu: Further, as per [databind#2522], passing `null` ValueInstantiator
        //   should simply trigger definition problem
        if (valueInst == null ) {
            msg = String.format("Cannot construct instance of %s: %s",
                    ClassUtil.nameOf(instClass), msg);
            return reportBadDefinition(instClass, msg);
        }
        if (!valueInst.canInstantiate()) {
            msg = String.format("Cannot construct instance of %s (no Creators, like default constructor, exist): %s",
                    ClassUtil.nameOf(instClass), msg);
            return reportBadDefinition(instClass, msg);
        }
        msg = String.format("Cannot construct instance of %s (although at least one Creator exists): %s",
                ClassUtil.nameOf(instClass), msg);
        return reportInputMismatch(instClass, msg);
    }